Shared vs own cloud kitchen: the white paper on the decision that sets your delivery margin
Verdict: in shared vs own cloud kitchen, SHARED wins when the brand has no proven demand yet and needs to validate zone, menu and average ticket with near-zero CapEx; OWN wins once the polygon sustains more than 55 to 60 daily orders per station and the variable-rent saving beats the amortization of the build-out. The tipping point is not the pride of owning a kitchen: it is the daily order count held steady for twelve weeks, measured by polygon rather than by city. With U.S. delivery concentrated in DoorDash at 60.7% of the market at year-end 2024 (Earnest Analytics, 2024) and the independent segment accounting for 61.7% of cloud kitchen revenue in 2025 (Grand View Research, 2025), an operator who decides without polygon-level data is betting CapEx against an algorithm they do not control.
An Asian food operator billing between 500 thousand and 1 million USD a year signed a two-year lease on an own kitchen in a polygon where the brand was doing 22 daily orders. The build-out cost what it cost and the fixed rent arrived punctually every month; the volume did not. That mistake has a technical name: installed capacity was bought before geolocated demand density was verified, and in delivery, density outranks the kitchen.
The shared versus own cloud kitchen debate is almost always framed as a rent comparison, and that is where the focus is lost. Rent is the visible number; the deciding number is total cost per order served, which absorbs aggregator commission, packaging, waste during peaks, last-mile cost when you run your own fleet and the opportunity cost of a station idle three hours a day. With Asia-Pacific reaching 48.0% of global cloud kitchen revenue in 2025 (Grand View Research, 2025), the model has matured enough for benchmarks to exist; ignoring them is a decision, not an accident.
This document approaches the decision from the local digital engine, which is where a virtual brand's real margin lives: local SEO, Google Business Profile, ranking algorithms across delivery platforms, geotargeted paid media by polygon and review reputation. An own kitchen badly placed relative to the demand centroid loses algorithmic visibility that no equipment investment recovers, because aggregators sort by distance and estimated time before perceived quality.
Side-by-side comparison
| SHARED cloud kitchen (third-party hub) | OWN cloud kitchen (turnkey build) | |
|---|---|---|
| Initial CapEx per station | ✕3,000 to 12,000 USD (mobile equipment, signage, POS and security deposit) | ✓45,000 to 140,000 USD (civil works, extraction, grease trap, permits and fixed equipment) |
| Occupancy cost structure | ✕Low fixed rent plus a variable fee of 8% to 15% on sales depending on contract | ✓Full fixed rent at 6% to 10% of sales plus building maintenance borne by the operator |
| Time to first invoiced order | ✕2 to 5 weeks (aggregator listing and local profile included) | ✓16 to 28 weeks (licensing, build-out, health inspection and commercial listing) |
| Break-even in daily orders | ✕28 to 40 orders/day per station at a 12 to 16 USD average ticket | ✓55 to 75 orders/day per station to amortize CapEx over 36 months |
| Control of the delivery polygon | ✕Limited: the location was chosen by the hub operator, not by your demand curve | ✓Total: you pick the centroid and optimize the estimated time that drives aggregator ranking |
| Virtual brand scalability | ✕High short term: adding a virtual brand costs the listing and the photography | ✓High long term: your own idle capacity absorbs brands with no incremental fee |
| Territory risk and exit | ✕Low: 3 to 12 month contracts, exit without sunk build-out | ✓High: 24 to 60 month terms and non-recoverable CapEx if the polygon underperforms |
| Exposure to aggregator concentration | ✕High: the hub usually imposes the channels integrated into its POS | ✓Medium: you can run a direct delivery channel and cut platform dependency |
Chapter 1 — Cost per order served, not rent, settles shared versus own
Compare total cost per order served, never rent against rent: that is where the whole decision sits. That figure adds aggregator commission, packaging, peak-hour waste, last mile if you run your own fleet, and the hours a station exists without producing. A shared kitchen turns almost all of it into a variable cost and charges you when you sell; an owned kitchen charges you for existing, whether orders come in or not. With Asia-Pacific holding 48.0% of global cloud kitchen revenue in 2025 (Grand View Research, 2025) and the independent segment taking 61.7% of that revenue, the model already has public benchmarks anyone can read before signing a two-year lease. An operator who compares only the monthly rent is measuring the part of the cost that moves least, and betting cash flow on the part that actually moves. The shared kitchen wins as long as you have not proven demand: unvalidated zone, unvalidated menu, unvalidated ticket, CapEx close to zero.
Chapter 2 — When does the shared kitchen win?
The case that opens this document says it plainly:
an Asian food brand doing 22 orders a day signed two years of its own build-out in an industrial zone, and the rent arrived on time every month while the volume never did. It bought installed capacity before checking geolocated demand density, and in delivery density beats kitchen. A shared hub lets you move the operation to another zone in 30 or 60 days, test three menus and kill two without touching your balance sheet. Mexico City went from its 2023 base to more than 1,200 active dark kitchens in 2025, a 40% jump (CANIRAC, 2025): there are stations available and hubs competing, which favors whoever negotiates without a build-out on their back. Move to your own kitchen when the zone sustains more than 55 to 60 daily orders per station steadily, not during a campaign spike.
Chapter 3 — The 55 to 60 daily orders per station threshold
That threshold is not arbitrary: below it, a hub's variable fee comes out cheaper than fixed rent plus equipment amortization; above it, every extra order entering an owned kitchen pays no incremental occupancy and the margin opens up. Verify it with four consecutive weeks of aggregator data, not with the best month of the quarter. Calculate labor separately, because in foodservice it runs between 25% and 35% of revenue according to the U.S. Bureau of Labor Statistics, and an owned kitchen carries all of it while a shared hub usually dilutes part of the support staff. If your curve crosses the threshold only on Fridays and Saturdays, you have crossed nothing yet. Every annual revenue band produces a different verdict, and forcing them all into one recommendation is the costliest mistake here. Below 500 thousand USD the answer is shared, no argument: any CapEx locks up the working capital that keeps the operation breathing.
Chapter 4 — The revenue band changes the answer
Between 500 thousand and 1 million —the band of the Asian operator in the case— shared still wins unless the zone already proves the threshold, because a two-year contract at that scale commits too much cash flow. Above 1 million the mixed model appears: an owned kitchen in the mature zone, shared hubs to explore new ones. Above 5 million an owned kitchen stops being a cost and becomes a portfolio platform, with several virtual brands sharing the same mise en place. Above 10 million the conversation is no longer shared versus own, but how many owned nodes and at what density. Above 5 million USD, a celebrity restaurant or a large-format themed concept is not buying a kitchen: it is buying brand control, and that carries its own cost structure. Shared kitchens almost always lose here, even when the initial volume does not justify the investment, because the asset at stake is reputation and a multi-brand hub cannot guarantee plating standard, supplier traceability or allergen protocol under your name.
Chapter 5 — High end: the celebrity chef restaurant and its own cost structure
The costs specific to this tier are different: a dedicated executive chef, menu R&D, quality control with documented audits, and signature packaging that can triple the unit cost of a generic box. With DoorDash Marketplace GOV growing +20% year over year during 2024 (DoorDash, full year 2024 results), the demand is there, but a brand that gets quoted in the press cannot let a third party improvise its product on a Friday at nine at night. The hub's location sets your estimated delivery time, and ETA reshuffles your position on Rappi or Uber Eats long before reviews do. Aggregators rank by distance and time before perceived quality within an urban radius, so an owned kitchen badly placed relative to the demand centroid loses visibility that no combi oven brings back. Shaving three minutes changes the ranking; going from 4.4 to 4.6 stars barely moves it. Concentration also matters when you negotiate: DoorDash closed 2024 with 60.7% of U.S.
Chapter 6 — Algorithmic visibility: ETA weighs more than half a star
delivery against 26.1% for Uber Eats and 6.3% for Grubhub (Earnest Analytics, 2024), which means in many markets you do not pick a platform, you pick a zone inside the dominant one. Choose that zone with the order heat map open, not with the property floor plan. A station idle for three hours costs you money in a shared kitchen and hands you options in your own: the same dead hour with the opposite sign. In a hub, those hours are fixed rent you pay without producing. In an owned kitchen, that same window is precisely the asset that lets you launch a second or third virtual brand with no incremental occupancy fee, spreading the same equipment across more tickets. There is the paradox that settles the decision: an owned kitchen costs more per order when you run one brand, and less when you run three. That is why an operator planning a portfolio should cross the threshold earlier than a single-brand operator.
Chapter 7 — Idle capacity flips sign depending on the model
Diego F. Parra and the Masterestaurant team work that transition on the local digital engine —Google Business Profile listing, zone-level paid media, aggregator ranking— because kitchens fill up from the map, not from the equipment. Run the scenario before you sign: if your dominant aggregator adds five points of commission in year two of an owned-kitchen lease, does your margin hold? With concentrations like iFood's 87% of Brazilian e-food bookings (Statista, 2024), Grab's 53.9% in Southeast Asia (Momentum Works, 2024), or over 95% of India split between Zomato and Swiggy (Business of Apps, 2025), pricing power is not on your side, and whoever carries fixed rent cannot negotiate with the door open. The operator in a shared hub trims the menu, cuts the weak zone and renegotiates; the one with an owned kitchen cuts spec sheets and quality, the worst lever available. This week, open your last four weeks of data by zone and calculate daily orders per station: if it does not reach 55, stay shared one more quarter.
Chapter 8 — The five differences operators discover too late
Shared charges you for selling; own charges you for existing. A weak month in a shared kitchen hurts the margin; a weak month in an own kitchen hurts the cash, because rent and amortization never consult your order curve before hitting the account. The shared hub picks your location and with it your estimated delivery time, which is the variable that weighs most in aggregator ranking within an urban radius. Shaving three minutes off the ETA reorders your position in the list long before half a review star does. Idle capacity flips sign. In a shared setup, a stopped station costs you that station's fixed rent. In your own kitchen, that same dead hour is the asset that makes launching a second or third virtual brand profitable with no incremental occupancy fee. The sales data is either yours or it is not. When the POS belongs to the hub and a third party administers the aggregator integrations, you lose the granularity for serious menu engineering, and without that granularity, contribution margin per dish gets estimated rather than measured.
Chapter 9 — The five differences operators discover too late — in practice
Exiting does not cost the same. Closing a shared station is a notice period; closing an own kitchen is sunk CapEx plus the remaining lease term, and with dark kitchens in Mexico City growing more than 40% since 2023 to surpass 1,200 active sites (CANIRAC, 2025), today's spare polygon can saturate tomorrow.
Criterion by criterion: where each model wins
When the SHARED kitchen is the financially correct callNear-zero CapEx
- Virtual brand with no demand history in the polygon: you are paying to learn, not to build.
- Projected annual revenue below 500 thousand USD in the unit's first year.
- Need to be selling within 45 days to capture a high season.
- Short menu of 14 to 22 SKUs that fits one station without specialized build-out.
- Simultaneous testing of two or three polygons before committing to a long lease.
- Leadership team with no prior experience managing health licensing and construction.
When the OWN kitchen stops being vanity and starts being marginMasterestaurant
- More than 55 daily orders sustained for twelve consecutive weeks in the same polygon.
- Three or more virtual brands sharing mise en place and absorbing idle capacity.
- Processes demanding fixed equipment: large convection oven, blast chiller, dedicated cold room.
- Multi-unit operation above 1 million USD annually with centralized purchasing.
- Need for a direct delivery channel to reduce aggregator fee dependency.
- Themed or celebrity-chef restaurant above 5 million USD producing for events and satellites.
Side-by-side comparison
| SHARED cloud kitchen (third-party hub) | OWN cloud kitchen (turnkey build) | |
|---|---|---|
| Initial CapEx per station | ✕3,000 to 12,000 USD (mobile equipment, signage, POS and security deposit) | ✓45,000 to 140,000 USD (civil works, extraction, grease trap, permits and fixed equipment) |
| Occupancy cost structure | ✕Low fixed rent plus a variable fee of 8% to 15% on sales depending on contract | ✓Full fixed rent at 6% to 10% of sales plus building maintenance borne by the operator |
| Time to first invoiced order | ✕2 to 5 weeks (aggregator listing and local profile included) | ✓16 to 28 weeks (licensing, build-out, health inspection and commercial listing) |
| Break-even in daily orders | ✕28 to 40 orders/day per station at a 12 to 16 USD average ticket | ✓55 to 75 orders/day per station to amortize CapEx over 36 months |
| Control of the delivery polygon | ✕Limited: the location was chosen by the hub operator, not by your demand curve | ✓Total: you pick the centroid and optimize the estimated time that drives aggregator ranking |
| Virtual brand scalability | ✕High short term: adding a virtual brand costs the listing and the photography | ✓High long term: your own idle capacity absorbs brands with no incremental fee |
| Territory risk and exit | ✕Low: 3 to 12 month contracts, exit without sunk build-out | ✓High: 24 to 60 month terms and non-recoverable CapEx if the polygon underperforms |
| Exposure to aggregator concentration | ✕High: the hub usually imposes the channels integrated into its POS | ✓Medium: you can run a direct delivery channel and cut platform dependency |
Market indicators framing the decision
“We closed the shared station in the north polygon and opened our own kitchen 1.8 km from the real order centroid. ETA dropped from 41 to 29 minutes, daily orders climbed from 38 to 71 across fourteen weeks and food cost settled at 29.4% once we centralized purchasing for three virtual brands into a single mise en place. The variable occupancy fee we paid the hub, 12% on sales, became fixed rent worth 7.1%: that gap of almost 5 points is what pays the 36-month amortization today.”
A 90-day roadmap to decide with data instead of instinct
Pull the order history by postal code from your aggregators and compute the weighted centroid of your demand, not the geographic average. Cross that point against the estimated delivery time each platform shows at peak hours, because ETA outweighs rating in list ordering within an urban radius. Document competitive density too: how many brands in your category occupy the first ten positions. With Uber Eats holding 26.1% of the U.S. market at year-end 2024 (Earnest Analytics, 2024), no single channel justifies a CapEx decision. Block output: a map with three candidate polygons and their estimated potential daily orders.
Model total cost per order served in shared and own setups using the same menu and the same average ticket. Include a food cost target below 32%, packaging, aggregator commission, labor cost inside the 25% to 35% of revenue range reported by the U.S. Bureau of Labor Statistics, and occupancy cost with its true nature: variable in the hub, fixed plus amortization in your own build. Compute break-even in daily orders for each scenario and the distance between your current volume and that threshold. Block output: two comparable financial models and one single number, the daily order count that makes the decision indifferent.
Before signing any build-out, run eight weeks from a shared station inside the candidate polygon and measure three things: sustained daily orders, 30-day repeat rate and average ticket variation between weekdays and weekends. Stand up the local digital engine in parallel, with a complete Google Business Profile, original product photography and a review request routine that holds the average above 4.6 stars. If the polygon fails to reach 45 daily orders by week eight, you have not found your zone yet. Block output: a documented decision to proceed, relocate or abort.
If the polygon responded, negotiate the own-kitchen lease with three concrete protections: stepped rent through the first six months, assignment rights and an early exit clause with a capped penalty. Present your board the 36-month ROI under three input-inflation scenarios plus the KPI you will report monthly, which is total cost per order served, not gross sales. With Europe holding 18.79% of the global dark kitchen market in 2024 (Global Growth Insights, 2024), international comparables exist and your board will ask for them. Block output: a signed contract with bounded risk and an active tracking dashboard.
And with AI?
Optimize channels, pricing and unit economics of your dark kitchen. Diego F. Parra is an expert in AI applied to restaurants.
Free tools to apply this now
Masterestaurant ecosystem tools for this decision
The shared versus own cloud kitchen call is not settled with a spreadsheet improvised the night before the board meeting. It needs a delivery unit economics model that separates variable cost from sunk cost, a cash projection that survives twelve weeks of slow ramp-up, and a map of the whole operation showing where each virtual brand's contribution margin actually lives.
The Masterestaurant framework Diego F. Parra applies across dark kitchen and foodtech puts those three pieces on the same board, so an own kitchen's CapEx is compared against a hub's variable fee under identical assumptions rather than uneven optimism.
Diego F. Parra — International consultant, expert in creating and scaling restaurants and in AI applied to restaurants, foodtech and HORECA. Methodology applied in 8.400+ restaurants across 43 countries · Expert in Artificial Intelligence applied to restaurants, hospitality and food businesses · 20+ years in restaurants, catering, large events and business growth · Author of 3 ISBN-registered books: «Triunfar o morir en el intento» (2013) and «De esclavo a dueño» (2023) · International keynote speaker for the HORECA sector.
Frequently asked questions about shared vs own cloud kitchen
How many daily orders justify building your own cloud kitchen?
How many daily orders justify building your own cloud kitchen?
The practical threshold sits between 55 and 75 daily orders per station sustained across twelve weeks, at a 12 to 16 USD average ticket. Below 45 daily orders, a shared hub's variable fee costs less than fixed rent plus 36-month build-out amortization.
Does an own kitchen improve my ranking on delivery apps?
Does an own kitchen improve my ranking on delivery apps?
Only if it brings the ETA closer to the customer. Aggregators sort by estimated time and distance within the urban radius before rating, so an own kitchen sited farther from the demand centroid than the previous hub worsens visibility despite costing far more.
Can I run several virtual brands from a shared kitchen?
Can I run several virtual brands from a shared kitchen?
Yes, and that is the model's most efficient use while you validate demand, though each additional brand usually pays its own occupancy fee. Once you reach three brands sharing mise en place, an own kitchen's idle capacity starts outperforming the hub.
What food cost should a dark kitchen operation hold?
What food cost should a dark kitchen operation hold?
The maximum tolerable is 32% per dish and the healthy delivery range runs between 27% and 30%, because packaging adds 2 to 4 points that dine-in never carries. Payroll, rent and utilities are not charged to the dish: they belong to break-even.
Sector data 2026 (official sources)
Verifiable industry benchmarks from official, non-commercial sources (government, industry associations, market research) - not competitors.
| Metric | Benchmark 2026 | Source |
|---|---|---|
| Cuota de Norteamérica en robótica de cocina | 40,8% | Grand View Research — Food Robotics Market |
| Mercado de robots de reparto en 2025 | USD 795,6 millones | MarketsandMarkets — Delivery Robots Market 2025 |
| Proyección del mercado de robots de reparto a 2030 | USD 3.236,5 millones (CAGR 32,4%) | MarketsandMarkets — Delivery Robots Market 2030 |
| Financiamiento de Starship Technologies en febrero de 2024 | USD 90 millones | Mordor Intelligence — Autonomous Delivery Robots Market |
| Entregas comerciales de Serve Robotics en Los Ángeles | >50.000 entregas | Serve Robotics — Form 8-K FY2024 (SEC) |
| Robots de Serve Robotics a desplegar en Uber Eats | hasta 2.000 robots | Serve Robotics — Form 8-K FY2024 (SEC) |
Download this document as PDF
The full text is free to read on this page. To take the corporate PDF with you, leave your details — we'll also email you the direct link.
Related content
Settle the decision before you sign
If you are weighing a shared versus own cloud kitchen and already hold order history by polygon, the next step is modeling total cost per order served under both scenarios with identical assumptions. The framework from Diego F. Parra and Masterestaurant turns that comparison into a single number your board can approve or reject without debating instincts.